CN113112516A - Image edge feature library construction method and device, computer equipment and storage medium - Google Patents

Abstract

The invention discloses an image edge feature library construction method, device, terminal equipment and storage medium. The method includes: transforming at least one edge image to be extracted by scale and/or angle to obtain an image set, where the image set includes multi-angle images and/or multi-scale images; After processing, a double threshold is determined according to the image set after thin edge processing; the double threshold includes a first threshold and a second threshold, and the first threshold is greater than the second threshold; according to the double threshold, each of the image sets is divided into The pixel points of the image are divided into the strong edge of the image, the weak edge of the image and the non-edge of the image, and the image set after edge extraction is obtained; according to the scale level of the image in the image set after the edge extraction, the image edge feature library is constructed to use for image edge extraction. This method can effectively solve the problem of weak self-adaptation in image edge contour extraction by adaptively calculating double thresholds according to the image.

Description

Image edge feature library construction method, device, computer equipment and storage medium

technical field

Embodiments of the present invention relate to the technical field of image processing, and in particular, to a method, apparatus, computer device, and storage medium for constructing an image edge feature library.

Background technique

With the development of digital processing technology, images have gradually become an important tool for obtaining information. The edge of the image contains high-frequency information in the image, such as road recognition of autonomous vehicles, face detection, etc., all rely on the edge detection and extraction of images. Image edge detection is the basis of image registration and target tracking, and is an important factor affecting the performance of the entire system.

In the prior art, the method of extracting image edges is to use quantum basic logic gates to establish 9 quantum image sets, calculate image gradient values through a quantum black box, classify the gradients by artificially setting thresholds, and then extract the edges of the quantum images. Due to the strong randomness of the artificially set gradient classification threshold, this method introduces the influence of artificial factors in the process of edge extraction, and the adaptive ability is weak.

Therefore, the technical problem of weak self-adaptation in image edge contour extraction is an urgent problem to be solved.

SUMMARY OF THE INVENTION

The embodiments of the present invention provide a method, device, computer equipment and storage medium for constructing an image edge feature library. The solution can effectively solve the problem of weak self-adaptation in image edge contour extraction by adaptively calculating double thresholds according to the image.

In a first aspect, an embodiment of the present invention provides a method for constructing an image edge feature library, including:

Perform scale and/or angle transformation on at least one edge image to be extracted to obtain an image set, where the image set includes multi-angle images and/or multi-scale images;

After performing thin edge processing on the image edge determined based on the image set, a double threshold value is determined according to the image set after the thin edge processing; wherein, the double threshold value includes a first threshold value and a second threshold value, and the first threshold value is greater than the second threshold value threshold;

Divide the pixel points of each image in the image set into strong edges of the image, weak edges of the image and non-edges of the image according to the double thresholds to obtain an image set after edge extraction;

An image edge feature library is constructed according to the scale level of the images in the image set after edge extraction, so as to be used for image edge extraction.

In a second aspect, the embodiment of the present invention also provides an image edge feature library construction device, including:

An image transformation module, configured to perform scale and/or angle transformation on at least one edge image to be extracted to obtain an image set, where the image set includes multi-angle images and/or multi-scale images;

A determination module, configured to perform thin edge processing on the image edge determined based on the image set and determine a double threshold value according to the image set after the thin edge processing; wherein, the double threshold value includes a first threshold value and a second threshold value, the first threshold value a threshold is greater than the second threshold;

a dividing module, configured to divide the pixels of each image in the image set into strong edges of the image, weak edges of the image and non-edges of the image according to the double thresholds, to obtain an image set after edge extraction;

The building module is configured to build an image edge feature library according to the scale level of the image in the image set after edge extraction, so as to be used for image edge extraction.

In a third aspect, an embodiment of the present invention also provides a computer device, including:

one or more processors;

a storage device for storing one or more programs;

The one or more programs are executed by the one or more processors, so that the one or more processors implement the image edge feature library construction method described in any embodiment of the present invention.

In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, implements the image edge feature library construction method provided by any embodiment of the present invention.

Embodiments of the present invention provide a method, device, computer equipment and storage medium for constructing an image edge feature library. First, at least one edge image to be extracted is scaled and/or angularly transformed to obtain an image set, where the image set includes multiple angles The image and/or multi-scale image; then the edge of the image determined based on the image set is subjected to thin edge processing and then a double threshold is determined according to the image set after the thin edge processing; then according to the double threshold value The pixels of each image are divided into the strong edge of the image, the weak edge of the image and the non-edge of the image, and the image set after edge extraction is obtained; finally, the image edge feature library is constructed according to the scale level of the image in the image set after edge extraction. , for image edge extraction. The above technical solution can effectively solve the problem of weak self-adaptation in image edge contour extraction by adaptively calculating dual thresholds according to the image.

Description of drawings

1 is a schematic flowchart of a method for constructing an image edge feature library according to Embodiment 1 of the present invention;

2 is a schematic flowchart of a method for constructing an image edge feature library according to Embodiment 2 of the present invention;

3 is a multi-angle image in a method for constructing an image edge feature library improved by Embodiment 2 of the present invention;

4 is a schematic diagram of an image set in a method for constructing an image edge feature library according to Embodiment 2 of the present invention;

5 is a schematic diagram of a directional difference image in a method for constructing an image edge feature library according to Embodiment 2 of the present invention;

6 is a schematic diagram of gradient direction division in a method for constructing an image edge feature library according to Embodiment 2 of the present invention;

7 is a schematic diagram of an image after thin edge processing in a method for constructing an image edge feature library according to Embodiment 2 of the present invention;

8 is a schematic diagram of an image after edge extraction in a method for constructing an image edge feature library according to Embodiment 2 of the present invention;

9 is a schematic diagram of an image set after edge extraction in a method for constructing an image edge feature library according to Embodiment 2 of the present invention;

10 is an overall flowchart of a method for constructing an image edge feature library according to Embodiment 2 of the present invention;

11 is a schematic structural diagram of an apparatus for constructing an image edge feature library according to Embodiment 3 of the present invention;

FIG. 12 is a schematic structural diagram of a computer device according to Embodiment 4 of the present invention.

Detailed ways

The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention. In addition, it should be noted that, for the convenience of description, the drawings only show some but not all structures related to the present invention.

Before discussing the exemplary embodiments in greater detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although a flowchart depicts various operations (or steps) as a sequential process, many of the operations may be performed in parallel, concurrently, or concurrently. Additionally, the order of operations can be rearranged. The process may be terminated when its operation is complete, but may also have additional steps not included in the figures. The processes may correspond to methods, functions, procedures, subroutines, subroutines, and the like. Furthermore, the embodiments of the invention and the features of the embodiments may be combined with each other without conflict.

The term "including" and its variants used in the present invention are open to include, ie, "including but not limited to". The term "based on" is "based at least in part on." The term "one embodiment" means "at least one embodiment."

Example 1

FIG. 1 is a schematic flowchart of a method for constructing an image edge feature library according to Embodiment 1 of the present invention. The method can be applied to a situation in which a large-scale edge feature library is constructed by extracting edge contours of an image. The method can be constructed from an image edge feature library. means, wherein the means may be implemented in software and/or hardware, and are typically integrated on computer equipment.

As shown in FIG. 1 , a method for constructing an image edge feature library provided in Embodiment 1 of the present invention includes the following steps:

S110. Perform scale and/or angle transformation on at least one edge image to be extracted to obtain an image set, where the image set includes multi-angle images and/or multi-scale images.

The edge image to be extracted may be a single-channel image, that is, a grayscale image. If the acquired image is a color image, the color image may be converted into a grayscale image, which is not specifically limited here.

The image set may be an image set composed of transformed images, and the image set may include multi-angle images, multi-scale images, and multi-angle and multi-scale images. The multi-angle images are multiple images with different angles obtained by transforming the edge images to be extracted at different angles; the multi-scale images are multiple images with different scales obtained after the edge images to be extracted are transformed by different scales; A multi-scale image is a multi-scale image with different scales and angles obtained by performing multi-angle transformation on the multi-scale image after the edge image to be extracted is scale-transformed to obtain a multi-scale image.

In this embodiment, the specific technical means for performing scale and/or angle transformation is not limited, as long as an image set can be obtained.

In one embodiment, the scale transformation of the edge image to be extracted may be to select different scale scaling factors within a preset scale range to scale the edge image to be extracted to obtain images with different scales.

In one embodiment, the method of performing angle transformation on the edge image to be extracted may be as follows: selecting the area to be rotated in the edge image to be extracted, and rotating the area to be rotated around the Z axis of the space coordinate system by constructing a space coordinate system, and according to the rotation Different angles can get multiple images with different angles.

In one embodiment, the method of performing scale and angle transformation on the edge image to be extracted may be as follows: first perform scale transformation on the edge image to be extracted to obtain a multi-scale image, and then perform angle transformation on the multi-scale image to obtain multiple images with different Angles and images at different scales.

The image set obtained in this step is a multi-scale image, a multi-angle image, and a multi-scale and multi-angle image, which can be used to construct an image edge feature library.

S120. After performing thin edge processing on the image edge determined based on the image set, determine a double threshold value according to the image set after the thin edge processing.

Wherein, the image edge may be the edge formed by the outline pixels of each image included in the image set; the double threshold may include a first threshold and a second threshold, and the first threshold is greater than the second threshold, and the first threshold may be one of the double thresholds The high threshold, the second threshold may be the low threshold of the dual thresholds. The first threshold and the second threshold can be adaptively calculated according to the information of the image.

In this embodiment, when the edge of the image in the image set is too thick, so that the edge information cannot be directly used to extract the image edge, it is necessary to suppress the pixels whose gradient amplitude is not large enough, and only need to keep the largest pixel. Gradient in order to achieve the purpose of thinning the edge.

The method of determining the image edge based on the image set may include performing edge detection on the images in the image set to obtain the image edge. It should be noted that the images in the image set need to be denoised before edge detection. The median filter method performs a certain degree of noise reduction on the image set. Among them, median filtering is a nonlinear digital filter technology, which is often used to remove noise in images or other signals, and has a good filtering effect on speckle noise and salt and pepper noise.

The process of median filtering is described in detail below. Exemplarily, median filtering replaces the value of a pixel with a grayscale median value in a predefined pixel neighborhood, that is:

表示(x,y)点的像素值，S_xy代表中心为(x,y)的8领域。in,

Represents the pixel value of the (x,y) point, and S _xy represents the 8-area centered at (x,y).

In this embodiment, the image after the thin edge processing will have noise points, that is, noise points, and the noise points are parts that are not required for edge extraction of the image, and therefore need to be filtered out. In view of the above situation, this embodiment uses double thresholds to acquire the strong edge of the image, the weak edge of the image, and the non-edge of the image, which can effectively remove the influence of noise. It should be noted that the double thresholds in this embodiment are adaptively calculated according to the image information, and the strong edges determined by the double thresholds have high accuracy.

In this embodiment, the thin edge processing may be non-maximum value suppression processing, and the image set after thin edge processing may be a gradient magnitude image set after non-maximum value suppression processing.

It should be further explained that the method of determining the double thresholds according to the image set after the thin edge processing may include: grading the images in the image set after the thin edge processing, and classifying the pixel values after the classification according to the first threshold value to be obtained and the The second threshold is divided into three categories, and the solved first and second thresholds are calculated by constructing an intra-class variance group.

S130. Divide the pixel points of each image in the image set into strong edges of the image, weak edges of the image, and non-edges of the image according to the double thresholds to obtain an image set after edge extraction.

Among them, the strong edge can be understood as the edge composed of strong pixels in the image, and it can also be understood as the clearer edge in the image; the weak edge can be understood as the edge composed of weak pixels in the image, and it can also be understood as the image. Blurred edges. Non-edge points can be understood as pixels in the image that do not constitute the outline of the image

It should be noted that the pixels with pixel values greater than the first threshold in the image can be regarded as strong pixels, the pixels with pixel values greater than the second threshold and less than the first threshold in the image can be regarded as weak pixels, and the pixel values in the image less than the first threshold can be regarded as weak pixels. Two-threshold points are considered non-edges.

In this embodiment, the edge contour point set can be obtained after the strong edge of the image, the weak edge of the image and the non-edge of the image are obtained. image set.

S140. Build an image edge feature library according to the scale level of the images in the image set after edge extraction, so as to be used for image edge extraction. In this embodiment, after obtaining the image set after edge extraction, an image edge feature library needs to be constructed according to the scale level of each image in the image set. The method for constructing the feature library is not specifically limited in this embodiment. .

In one embodiment, image edge extraction can be performed according to an image edge feature library. When the edge contour of an image needs to be extracted, an edge feature library can be constructed according to the image, and then the image can be compared with the image edge feature library in the image edge feature library. The edge-extracted image set is compared and matched, and then the edge-extracted image of the image can be output.

In a method for constructing an image edge feature library provided by the first embodiment of the present invention, at least one edge image to be extracted is firstly scaled and/or angularly transformed to obtain an image set, where the image set includes multi-angle images and/or multi-scale images secondly, after thin edge processing is performed on the edge of the image determined based on the image set, a double threshold is determined according to the image set after thin edge processing; then the pixel points of each image in the image set are divided according to the double threshold It is divided into strong edges of the image, weak edges of the image and non-edges of the image to obtain an image set after edge extraction; finally, an image edge feature library is constructed according to the scale level of the image in the image set after the edge extraction, so as to be used for image processing. edge extraction. The above method does not need to set double thresholds, and can adaptively calculate the thresholds according to the image, which can solve the problem of redundant or too few edge points caused by factors that are considered to be empirically set, and improve the efficiency of edge extraction.

Embodiment 2

FIG. 2 is a schematic flowchart of a method for constructing an image edge feature library according to Embodiment 2 of the present invention. This Embodiment 2 is optimized on the basis of the foregoing embodiments. In this embodiment, after performing thin edge processing on the image edge determined based on the image set, double thresholds are determined according to the image set after thin edge processing for further optimization; The threshold divides the pixel points of each image in the image set into strong edges of the image, weak edges of the image and non-edges of the image, to obtain an image set after edge extraction for further optimization. Please refer to the first embodiment for the content that is not yet detailed in this embodiment.

As shown in FIG. 2 , a method for constructing an image edge feature library provided in Embodiment 2 of the present invention includes the following steps:

S210. Perform scale and/or angle transformation on at least one edge image to be extracted to obtain an image set, where the image set includes multi-angle images and/or multi-scale images.

In this step, obtaining an image set may include: performing scale transformation on at least one edge image to be extracted to obtain an image set, performing angle transformation on at least one edge image to be extracted to obtain an image set, and scaling at least one edge image to be extracted The angle transformation results in an image set.

Further, perform scale transformation on at least one edge image to be extracted to obtain an image set, including:

Step 1. Determine a plurality of different scale scaling factors according to the preset image scale range and scale step size.

The scaling factor of the image in the X and Y directions can be calculated according to the preset image scale range and scale step, and the scaling factor can determine the scaling ratio of the image in the X and Y directions.

Exemplarily, according to the preset image scale range [scale _min , scale _max ] and the scale transformation step scalestep, the scale scaling factor of the edge image to be extracted in the X and Y directions can be calculated according to the following formula:

Among them, scale _{factor_x} and scale _{factor_y} represent the scale factor of the image in the X and Y directions, respectively. Exemplarily, the preset scale range may be [0.5, 1.5], and the scale step may be 0.1.

Step 2: Scale at least one edge image to be extracted according to the scale scaling factor to obtain an image set.

In this step, within the preset image scale range, different scale scaling factors can be selected according to the scale step to perform different scale scaling on the edge image to be extracted, so as to obtain multiple multi-scale images with different scales.

Further, performing angle transformation or scale angle transformation on at least one edge image to be extracted to obtain an image set, including: establishing a spatial coordinate system according to the to-be-transformed image; The image set is obtained by rotating around the Z axis of the space coordinate system.

The image to be transformed may include at least one edge image to be extracted or a multi-scale image.

In this step, when the image to be transformed includes at least one edge image to be extracted, a spatial coordinate system is established according to the center point of the edge image to be extracted as the origin, and the area to be rotated in the image to be extracted is rotated around the The Z-axis of the space coordinate system is rotated to obtain a multi-angle image.

In this step, when the image to be transformed includes at least one multi-scale image, a spatial coordinate system is established according to the center point of the scale image as the origin, and the area to be rotated in the scale image is rotated around the spatial coordinates according to a preset rotation angle. The Z axis of the system is rotated to obtain multi-angle images.

Further, rotating the area to be rotated in the to-be-transformed image around the Z-axis of the spatial coordinate system according to a preset rotation angle to obtain an image set, including: determining a rotation matrix based on a preset rotation angle; and the rotation center coordinates of the image to be transformed to determine a transformation matrix; determine the minimum inscribed circle radius of the to-be-transformed image according to the rotation center coordinates; determine the to-be-rotated area based on the minimum inscribed circle radius; The pixel coordinates in the area other than the to-be-rotated area in the transformed image are taken as static coordinates; the angle-transformed pixel coordinates are determined according to the transformation matrix and the pixel coordinates in the to-be-rotated area; according to the static coordinates And the pixel coordinates after angle transformation to get the image set.

The preset rotation angle angleStart may be the preset rotation angle of the image to be transformed around the Z-axis of the space coordinate system, the rotation matrix may be the rotation matrix of affine transformation, and the formula for calculating the rotation matrix is:

Among them, θ represents the preset rotation angle, and Z _rot(θ) represents the rotation matrix corresponding to the rotation angle θ of the image to be transformed.

In this embodiment, the rotation center coordinate can be understood as the coordinate center of the image to be transformed, and half of the row pixel value and half of the column pixel value of the to-be-transformed image can be taken as the rotation center coordinate. Exemplarily, if the row pixel value is cols and the column pixel value is rows, then the coordinates of the rotation center may be (cols/2,rows/2).

Among them, the transformation matrix can be a radiation transformation matrix containing the rotation and translation information of the image to be transformed, and its expression is as follows:

It should be noted that each value of θ _i is the same, that is, the angle step angleStep of each rotation of the image to be transformed is the same.

Since the image has the problem of edge filling after rotation, rotating according to the minimum inscribed circle of the image to be transformed can ensure that the edge of the rotated image is still valid.

Wherein, the minimum inscribed circle radius may be the minimum value of the abscissa and ordinate coordinates of the center coordinates. Exemplarily, the rotation center coordinates are (cols/2,rows/2), then the minimum inscribed circle radius is r=Min(cols /2,rows/2).

The area to be rotated may be a circular area, and the radius of the circle is the radius of the minimum inscribed circle.

In this embodiment, the method of determining the pixel coordinates after angle transformation includes: traversing the pixel rows of the image to be transformed, and obtaining the pixel coordinates after angle transformation according to the pixel coordinates of the image to be transformed in the area to be rotated and the transformation matrix.

Exemplarily, the formula for calculating the pixel coordinates after angle transformation is:

Among them, _ui represents the abscissa of the transformed pixel coordinates obtained after the ith rotation of the image to be transformed, v _i represents the ordinate of the transformed pixel coordinates obtained after the ith rotation of the image to be transformed, and _xi represents the ordinate of the transformed pixel coordinates. The abscissa of the pixel coordinates of the transformed image in the area to be rotated, and y _i represents the ordinate of the pixel coordinates of the to-be-transformed image in the area to be rotated. According to the above formula, the radiation-transformed pixel coordinates of the image to be transformed in the area to be transformed can be calculated, so that the rotation of the image to be transformed can be realized. FIG. 3 is a multi-angle image in a method for constructing an image edge feature library improved by Embodiment 2 of the present invention. As shown in FIG. 3 , a multi-angle image can be obtained after an image to be transformed is rotated in the area to be rotated Image.

In this embodiment, only the to-be-rotated area in the to-be-transformed image needs to be rotated, and other areas of the to-be-transformed image maintain the original angle, that is, the pixel coordinates in the static area in the to-be-transformed image are taken as static coordinates and remain unchanged. Therefore, a multi-angle image includes stationary coordinates and angle-transformed pixel coordinates.

FIG. 4 is a schematic diagram of an image set in a method for constructing an image edge feature library according to Embodiment 2 of the present invention. As shown in Figure 4, the nine images in the image set have different scales and different angles, respectively.

S220. Perform a direction difference on each image in the image set to obtain a direction difference image.

Before this step, it is also necessary to perform noise reduction processing on each image in the image set, that is, noise reduction through fast median filtering. The specific process has been described in Embodiment 1 and will not be repeated here.

The directional difference may be the Sobel image directional difference. In image processing, the Sobel image directional difference is essentially a first-order filter.

Exemplarily, the convolution kernel of the Sobel operator may include two sets of 3×3 matrices, which are the convolution kernel in the x-direction and the convolution kernel in the y-direction, respectively, and the expressions are as follows:

Through the convolution kernel in the x direction and the convolution kernel in the y direction, the image set after noise reduction can be convolved, and the direction difference image can be obtained. The difference image can be calculated by the following formula:

Among them, S _x and S _y are the convolution kernels in the x and y directions, respectively, L(x, y) is the pixel coordinates of the image in the image set after noise reduction, G _x (x, y) and G _y (x , y) are the difference value in the x direction and the difference value in the y direction, respectively. A differential image can be obtained according to its differential value and its corresponding pixel coordinates.

5 is a schematic diagram of a direction difference image in a method for constructing an image edge feature library according to Embodiment 2 of the present invention. As shown in FIG. 5 , the direction difference image may include an x-direction difference image and a y-direction difference image.

S230 , performing a non-maximum value suppression process on the directional difference image to obtain a thin-edge processed image set.

In this step, the gradient magnitude of the directional difference image needs to be calculated first, and the gradient magnitude of the corresponding image can be calculated according to the difference value in the x direction and the difference value in the y direction. The calculation formula is as follows:

Among them, F(x,y) is the gradient magnitude of the corresponding coordinate.

Further, the gradient direction can be calculated according to the gradient magnitude, and the calculation formula is:

θ=atan2(G _y (x,y),G _x (x,y))

Among them, θ is the gradient direction of the corresponding coordinate.

In this embodiment, after the gradient directions are obtained, the central pixel point of the differential image can be divided into four directions according to the gradient angles. FIG. 6 is a schematic diagram of gradient direction division in a method for constructing an image edge feature library according to Embodiment 2 of the present invention. As shown in FIG. 6 , a differential image can be divided into four directions by four straight lines.

Exemplarily, the differential image can be divided into the following four directions:

Horizontal gradient direction: (θ>-22.5&&θ<22.5)||(θ<-157.5&&θ>157.5) That is, the double arrow direction between straight line 3 and straight line 4 in Figure 6;

Vertical gradient direction: (θ>=67.5&&θ<=112.5)||(θ>=-112.5&&θ<=-67.5) That is, the double arrow direction between straight line 1 and straight line 2 in Figure 6;

The upper right diagonal gradient direction: (θ>=22.5&&θ<67.5)||(θ>=-157.5&&θ<-112.5), which is the double arrow direction between straight line 2 and straight line 4 in Figure 6;

The lower right diagonal gradient direction: (θ>=-67.5&&θ<=-22.5)||(θ>112.5&&θ<=157.5), which is the double arrow direction between straight line 1 and straight line 3 in FIG. 6 .

In this embodiment, after the four directions are obtained, the gradient values of the pixel points on the gradient line in the 8 domains can be compared respectively. Among them, the 8 areas are divided into 8 adjacent areas by the straight line 1, the straight line 2, the straight line 3 and the straight line 4, the gradient straight line can be the four double-arrow straight lines in FIG. 6, and the gradient straight line can include the horizontal gradient straight line, Vertical gradient line, lower left upper right -45° gradient line, and upper left right lower 45° gradient line.

Further, according to the obtained gradient direction of the corresponding coordinate, it can be determined which direction it is in the above four directions. Exemplarily, if the gradient direction of the corresponding coordinate is θ=80°, it can be determined that it belongs to the vertical gradient direction. After the direction is determined, the gradient line in the direction can be determined, and the gradient value of the center point and the gradient values of the two endpoints on the gradient line can be calculated. When the gradient value of the center point is greater than the gradient value of the two endpoints on the gradient line, the The gradient value of the center point is determined as the maximum value in the 8 fields, and the pixel value corresponding to the center point is taken as the selected pixel; when the gradient value of the center point is less than or equal to the gradient value of the two endpoints on the gradient line, the center point The pixel value corresponding to the point is 0. In this way, the contour edge of the directional difference image can be thinned to obtain a thinned image set. FIG. 7 is a schematic diagram of an image after thin edge processing in a method for constructing an image edge feature library according to Embodiment 2 of the present invention.

S240. Determine pixel classification according to the image set after the thin edge processing.

Among them, pixel classification can be understood as dividing an image into multiple levels according to pixel values.

This step may include grading the pixel values of the images according to the number of image bits in the images in the thin edge processed image set. The images in the image after the thin edge processing are classified according to the pixel value of each image, and the total number of classifications can be determined according to the number of bits of the image. Divided into [0,1,...,255], the total number of classifications L=256.

S250. Determine dual thresholds according to the pixel classification.

In this embodiment, according to the pixel classification, the pixel values of the image after the thin edge processing can be divided into three categories according to the first threshold to be obtained and the second threshold to be obtained. A threshold value and a second threshold value to be obtained are solved to obtain a double threshold value.

Further, this step may include the following steps:

S251, classifying the graded pixel values into a non-edge point pixel value class, a weak edge point pixel value class and a strong edge point pixel value class according to the first threshold value to be obtained and the second threshold value to be obtained;

Wherein, the first threshold value to be determined may be a value for segmenting the pixel value class of strong edges, and the second threshold value to be determined may be a value for segmenting the pixel value class of weak edge edges. Exemplarily, the first threshold to be determined may be m, and the second threshold to be determined may be k.

Exemplarily, the pixel values in the graded image are further divided into three categories: C ₀ , C ₁ and C ₂ , where C ₀ =[0,1,...,k] represents the non-edge point pixel value category, and C ₁ =[k,k+1,...,m] represents the weak edge point pixel value class, and C ₂ =[m+1,m+2,...,L-1] represents the strong edge point pixel value class.

S252. Construct non-edge point pixel value class variance, weak edge point pixel value class variance, and strong edge point pixel value class variance according to the graded pixel values, the first threshold value to be determined, and the second threshold value to be determined, to obtain the intra-class variance variance group;

In this step, the proportion of each level of pixels in the entire thin-edge processed image can be calculated according to the pixel classification, and the calculation formula is as follows:

Among them, P _i represents the proportion of i-th level pixels in the whole image, n _i is the number of i-th level pixels, and N is the total number of pixels in the image.

Exemplarily, the value of the first threshold to be determined is m, and the value of the second threshold to be determined is k. The sum of the pixel probabilities of the top k items less than the second threshold to be determined can be calculated, and the calculation formula is as follows:

where ω ₀ (k) represents the sum of the pixel probabilities of the first k items.

Based on the pixel probability sum of the top k items, the expected sum of the top k items less than the second threshold to be calculated can be further calculated. The calculation formula is as follows:

where μ ₀ (k) represents the expected sum of the first k terms.

It is also possible to calculate the probability sum that is higher than the second threshold to be determined and less than the first threshold to be determined, and the calculation formula is as follows:

Wherein, ω ₁ (k,m) represents the probability sum that is higher than the second threshold to be determined and smaller than the first threshold to be determined.

Based on the probability sum that is higher than the second threshold to be determined and smaller than the first threshold to be determined, the expected sum that is higher than the second threshold to be determined and smaller than the first threshold to be determined can be further calculated. The calculation formula is as follows:

Wherein, μ ₁ (k,m) represents the expected sum that is higher than the second threshold to be determined and smaller than the first threshold to be determined.

You can also continue to calculate the probability sum that is higher than the first threshold to be calculated. The formula is as follows:

Among them, ω ₂ (m) represents the probability sum that is higher than the first threshold value to be obtained.

Based on the probability sum higher than the first threshold to be calculated, the expected sum higher than the first threshold can be further calculated, and the calculation formula is as follows:

where μ ₂ (m) represents the expected sum above the first threshold.

Based on the parameters calculated above, the intra-class variance group can be further constructed, and the constructed intra-class variance group is:

表示非边缘点像素值类方差，

表示弱边缘点像素值类方差，

表示强边缘点像素值类方差。in,

represents the class variance of pixel values of non-edge points,

represents the class variance of weak edge point pixel values,

Represents the class variance of pixel values of strong edge points.

S253, establishing a minimization evaluation function according to the intra-class variance group, and obtaining a partial derivative of the minimization evaluation function to obtain a minimization calculation formula;

In this step, the minimum evaluation function can be calculated based on the intra-class variance group constructed above, and the calculation formula is:

Among them, J(k,m) represents the minimization evaluation function. In order to minimize it, the partial derivative of the minimized evaluation function needs to be obtained to obtain the minimized formula:

According to the above formula, the minimization formula can be obtained.

S254 , traversing the to-be-determined first threshold value and the to-be-determined second threshold value within the graded pixel values to obtain the solved first threshold value and the second threshold value that satisfy the minimization calculation formula.

In this step, the solved first threshold and the first threshold are dual thresholds.

S260: Determine the pixel points higher than the first threshold in each image in the image set as strong edges, determine the pixels higher than the second threshold and lower than the first threshold as weak edges, and determine the pixels lower than the second threshold as weak edges. Pixels are determined to be non-edges.

In this embodiment, this step belongs to hysteresis threshold processing. According to the obtained double thresholds, it can be set that the pixels higher than the first threshold belong to strong edges, and the pixels lower than the first threshold and higher than the second threshold belong to weak edges. Pixels below the second threshold are classified as non-edges.

Further, after obtaining the weak edge, it also includes: determining whether each pixel in the area formed by the weak edge has a strong edge pixel in a preset numerical neighborhood; The edge pixel is set as a strong edge; if it does not exist, the pixel is removed.

Among them, in the weak edge part, it is judged whether there is a strong edge pixel point for each pixel point within 8 fields, if there is, the pixel point is reset and divided into a strong edge point; if there is no strong edge pixel point, the pixel point is Click to remove. This operation removes individual noise in the image and also enlarges strong edges.

FIG. 8 is a schematic diagram of an image after edge extraction in a method for constructing an image edge feature library according to Embodiment 2 of the present invention. As shown in FIG. 8 , the first threshold value of the image adaptively calculated is 150, and the second threshold value is 92, and the image is obtained by hysteresis threshold processing according to the double thresholds.

S270. Perform edge extraction on the image set based on the strong edge, weak edge, and non-edge to obtain an edge-extracted image set.

In this step, the non-edge part of the image in the image set is not subjected to edge extraction, the pixel value of the weak edge part of the image in the image set is set to 0, and the strong edge of the image in the image set is extracted into the edge to obtain the image set after edge extraction. .

FIG. 9 is a schematic diagram of an image set after edge extraction in a method for constructing an image edge feature library according to Embodiment 2 of the present invention. As shown in FIG. 9 , the image set after edge extraction may include multiple images with different scales, Image after edge extraction at different angles and different scale angles.

S280. Build an image edge feature library according to the scale level of the images in the image set after edge extraction, so as to be used for image edge extraction.

In this embodiment, the edge points of a single image in the image set after edge extraction are stored in a continuous memory, and an unordered map data structure is used for the scale level. Exemplarily, the image edge data structure may be: unordered_map<float ,vector<vector<Points>>>>, where float represents the scale layer, which represents the edge contour point set containing the rotation angle.

The second embodiment of the present invention provides a method for constructing an image edge feature library. The method is not limited to the edge extraction of a certain angle or a certain scale image, and can realize the construction of a multi-scale and multi-angle image set for the same edge image to be extracted. , which provides an effective guidance strategy for building a large-scale image edge feature library. In addition, the method can adaptively calculate double thresholds according to image information for image edge extraction to improve edge extraction efficiency. The edge extraction in this method adopts the data structure of disordered graph and continuous memory storage. Through nesting, multi-level large-scale image edge feature information storage can be realized, and the complexity of the data structure can be reduced.

Embodiment 2 of the present invention also provides an overall flow chart of a method for constructing an image edge feature library. FIG. 10 is an overall flow chart of a method for constructing an image edge feature library according to Embodiment 2 of the present invention. As shown in FIG. 10 , a The overall process of the image edge feature library construction method can include the following processes:

Obtain the edge image to be extracted, perform multi-scale and multi-angle transformation on the edge image to be extracted to obtain an image set; perform fast median filtering on the images in the image set, calculate the Sobel image direction difference, calculate the image gradient magnitude and direction, and compare the direction difference image. Perform non-maximum suppression; establish image gradient amplitude histogram, that is, the image after thin edge processing, construct intra-class variance group, calculate double threshold, and finally perform lag threshold processing and output the edge data set to be extracted to obtain the image edge feature library.

Embodiment 3

11 is a schematic structural diagram of an apparatus for constructing an image edge feature library according to Embodiment 3 of the present invention. The apparatus is applicable to the case of extracting the edge contour of an image to construct a large-scale edge feature library, wherein the apparatus can be configured by software and/or hardware implemented and generally integrated on computer equipment.

As shown in FIG. 11 , the apparatus includes: an image transformation module 111 , a determination module 112 , a division module 113 and a construction module 114 .

An image transformation module 111, configured to perform scale and/or angle transformation on at least one edge image to be extracted to obtain an image set, where the image set includes multi-angle images and/or multi-scale images;

The determining module 112 is configured to perform thin edge processing on the image edge determined based on the image set and determine a double threshold value according to the image set after the thin edge processing; wherein, the double threshold value includes a first threshold value and a second threshold value, the the first threshold is greater than the second threshold;

The dividing module 113 is used to divide the pixel points of each image in the image set into strong edges of the image, weak edges of the image and non-edges of the image according to the double threshold, to obtain the image set after edge extraction;

The construction module 114 is configured to construct an image edge feature library according to the scale level of the image in the image set after the edge extraction, so as to be used for image edge extraction.

In this embodiment, the apparatus first performs scale and/or angle transformation on at least one edge image to be extracted through the image transformation module to obtain an image set, the image set includes multi-angle images and/or multi-scale images; secondly The image edge determined based on the image set is subjected to thin edge processing by the determining module, and a double threshold is determined according to the image set after the thin edge processing; then the pixel of each image in the image set is divided by the dividing module according to the double threshold The points are divided into the strong edge of the image, the weak edge of the image and the non-edge of the image, and an image set after edge extraction is obtained; Used for image edge extraction. The device can effectively solve the problem of weak self-adaptation in image edge contour extraction by adaptively calculating double thresholds according to the image.

Further, the image transformation module 111 includes: a scale transformation unit and a scale angle transformation unit.

The scale transformation unit is configured to determine a plurality of different scale scaling factors according to the preset image scale range and scale step size; scale at least one edge image to be extracted according to the scale scaling factors to obtain an image set, the The image set includes images of multiple scales.

The scale and angle transformation unit is used to establish a spatial coordinate system according to the image to be transformed; rotate the area to be rotated in the image to be transformed around the Z axis of the spatial coordinate system according to a preset rotation angle to obtain an image set; The image to be transformed includes at least one edge image to be extracted or a multi-scale image; when the image to be transformed includes at least one edge image to be extracted, the image set includes multi-angle images; When including at least one multi-scale image, the image set includes multi-angle and multi-scale images.

On the basis of the above optimization, the scale and angle transformation unit is further configured to determine a rotation matrix based on a preset rotation angle; determine a transformation matrix according to the rotation matrix and the rotation center coordinates of the to-be-transformed image; determine the rotation matrix according to the rotation center coordinates The minimum inscribed circle radius of the image to be transformed; the area to be rotated is determined based on the minimum inscribed circle radius; the coordinates of pixels in the area other than the area to be rotated in the image to be transformed are taken as static coordinates; The transformation matrix and the coordinates of the pixel points in the area to be rotated determine the pixel coordinates after angle transformation; an image set is obtained according to the static coordinates and the pixel coordinates after angle transformation.

Based on the above technical solution, the determining module 112 is specifically configured to: perform directional difference on each image in the image set to obtain a directional difference image; perform non-maximum suppression processing on the directional difference image to obtain a thin edge processed image set ; determining a pixel classification according to the image set after the thin edge processing; determining a double threshold value according to the pixel classification.

Further, the determining module 112 is further configured to: classify the pixel values of the images according to the number of bits of the images in the image set after the thin edge processing; classify the pixel values after the classification according to the first threshold to be determined and the first The two thresholds are divided into non-edge point pixel value class, weak edge point pixel value class and strong edge point pixel value class; non-edge point pixels are constructed according to the classified pixel values, the first threshold value to be determined and the second threshold value to be determined value class variance, weak edge point pixel value class variance, and strong edge point pixel value class variance to obtain an intra-class variance group; establish a minimized evaluation function according to the intra-class variance group, and obtain the partial derivative of the minimized evaluation function to obtain Minimizing the calculation formula; traversing the to-be-calculated first threshold and the to-be-calculated second threshold within the graded pixel values to obtain the solved first and second thresholds that satisfy the minimization calculation formula.

Further, the dividing module 113 is specifically configured to determine the pixel points higher than the first threshold in each image in the image set as strong edges, and determine the pixels higher than the second threshold and lower than the first threshold as weak edges, Determine the pixel points below the second threshold as non-edges; perform edge extraction on the image set based on the strong edges, weak edges and non-edges to obtain an edge-extracted image set.

Further, after obtaining the weak edge, the dividing module 113 is also used to determine whether each pixel in the region formed by the weak edge has a strong edge pixel in the preset numerical neighborhood; The strong edge pixel is set as a strong edge; if it does not exist, the pixel is removed.

The above image edge feature library construction apparatus can execute the image edge feature library construction method provided by any embodiment of the present invention, and has functional modules and beneficial effects corresponding to the execution method.

Embodiment 4

FIG. 12 is a schematic structural diagram of a computer device according to Embodiment 4 of the present invention. As shown in FIG. 12 , the computer device provided by the fourth embodiment of the present invention includes: one or more processors 121 and a storage device 122; the processor 121 in the computer device may be one or more, and in FIG. 12, one processor Take the processor 121 as an example; the storage device 122 is used to store one or more programs; the one or more programs are executed by the one or more processors 121, so that the one or more processors 121 implement the present invention The image edge feature library construction method according to any one of the embodiments.

The computer equipment may further include: an input device 123 and an output device 124 .

The processor 121 , the storage device 122 , the input device 123 , and the output device 124 in the computer equipment may be connected by a bus or in other ways, and the connection by a bus is taken as an example in FIG. 12 .

As a computer-readable storage medium, the storage device 122 in the computer device may be used to store one or more programs, and the programs may be software programs, computer-executable programs, and modules, as described in Embodiment 1 or Embodiment 2 of the present invention. The program instructions/modules corresponding to the provided image edge feature library construction method (for example, the modules in the image edge feature library construction device shown in FIG. 114). The processor 121 executes various functional applications and data processing of the computer equipment by running the software programs, instructions and modules stored in the storage device 122, ie, implements the image edge feature library construction method in the above method embodiments.

The storage device 122 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the computer equipment, and the like. Additionally, storage device 122 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some examples, storage device 122 may further include memory located remotely from processor 121, which may be connected to the device through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

The input device 123 may be used to receive input numerical or character information, and to generate key signal input related to user settings and function control of the computer device. The output device 124 may include a display device such as a display screen.

Moreover, when one or more programs included in the above-mentioned computer device are executed by the one or more processors 121, the programs perform the following operations:

Perform scale and/or angle transformation on at least one edge image to be extracted to obtain an image set, where the image set includes multi-angle images and/or multi-scale images;

After performing thin edge processing on the image edge determined based on the image set, a double threshold is determined according to the image set after the thin edge processing; wherein, the double threshold includes a first threshold and a second threshold, and the first threshold is greater than the first threshold. two thresholds;

Divide the pixel points of each image in the image set into strong edges of the image, weak edges of the image and non-edges of the image according to the double thresholds to obtain an image set after edge extraction;

An image edge feature library is constructed according to the scale level of the images in the image set after edge extraction, so as to be used for image edge extraction.

Embodiment 5

Embodiment 5 of the present invention provides a computer-readable storage medium on which a computer program is stored, and when the program is executed by a processor, is used to execute a method for constructing an image edge feature library, and the method includes:

Perform scale and/or angle transformation on at least one edge image to be extracted to obtain an image set, where the image set includes multi-angle images and/or multi-scale images;

After performing thin edge processing on the image edge determined based on the image set, a double threshold is determined according to the image set after the thin edge processing; wherein, the double threshold includes a first threshold and a second threshold, and the first threshold is greater than the first threshold. two thresholds;

Divide the pixel points of each image in the image set into strong edges of the image, weak edges of the image and non-edges of the image according to the double thresholds to obtain an image set after edge extraction;

An image edge feature library is constructed according to the scale level of the images in the image set after edge extraction, so as to be used for image edge extraction.

Optionally, when the program is executed by the processor, it can also be used to execute the image edge feature library construction method provided by any embodiment of the present invention.

The computer storage medium in the embodiments of the present invention may adopt any combination of one or more computer-readable mediums. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination of the above. More specific examples (non-exhaustive list) of computer readable storage media include: electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read only memory (Read Only Memory, ROM), Erasable Programmable Read Only Memory (Erasable Programmable Read Only Memory, EPROM), flash memory, optical fiber, portable CD-ROM, optical storage device, magnetic storage device, or any suitable combination of the above. A computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus, or device.

A computer-readable signal medium may include a propagated data signal in baseband or as part of a carrier wave, with computer-readable program code embodied thereon. Such propagated data signals may take a variety of forms including, but not limited to, electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device .

Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber cable, radio frequency (RF), etc., or any suitable combination of the foregoing.

Computer program code for carrying out operations of the present invention may be written in one or more programming languages, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional A procedural programming language, such as the "C" language or similar programming language. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any kind of network, including a local area network (LAN) or wide area network (WAN), or may be connected to an external computer (eg, through the Internet using an Internet service provider) connect).

Note that the above are only preferred embodiments of the present invention and applied technical principles. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and can also include more other equivalent embodiments without departing from the concept of the present invention. The scope is determined by the scope of the appended claims.

Claims (11)

1. An image edge feature library construction method is characterized by comprising the following steps:

carrying out scale and/or angle transformation on at least one edge image to be extracted to obtain an image set, wherein the image set comprises multi-angle images and/or multi-scale images;

performing thin edge processing on the image edge determined based on the image set, and determining a double threshold value according to the image set subjected to thin edge processing; wherein the dual threshold comprises a first threshold and a second threshold, the first threshold being greater than the second threshold;

dividing pixel points of each image in the image set into a strong edge of the image, a weak edge of the image and a non-edge of the image according to the double thresholds to obtain an image set after edge extraction;

and constructing an image edge feature library according to the scale levels of the images in the image set after the edge extraction so as to extract the image edges.

2. The method according to claim 1, wherein the scaling at least one edge image to be extracted to obtain an image set comprises:

determining a plurality of different scale scaling factors according to a preset image scale range and a scale step length;

and carrying out scale scaling on at least one edge image to be extracted according to the scale scaling factor to obtain an image set, wherein the image set comprises multi-scale images.

3. The method according to claim 1, wherein the subjecting at least one edge image to be extracted to angle transformation or scale-angle transformation to obtain an image set comprises:

establishing a space coordinate system according to the image to be transformed;

rotating the region to be rotated in the image to be transformed around the Z axis of the space coordinate system according to a preset rotation angle to obtain an image set;

the image to be transformed comprises at least one edge image to be extracted or a multi-scale image;

when the image to be transformed comprises at least one edge image to be extracted, the image set comprises images of multiple angles; when the image to be transformed comprises at least one multi-scale image, the image set comprises multi-angle multi-scale images.

4. The method according to claim 3, wherein the rotating the region to be rotated in the image to be transformed around the Z axis of the spatial coordinate system according to a preset rotation angle to obtain an image set comprises:

determining a rotation matrix based on a preset rotation angle;

determining a transformation matrix according to the rotation matrix and the rotation center coordinate of the image to be transformed;

determining the minimum inscribed circle radius of the image to be transformed according to the rotation center coordinate;

determining a region to be rotated based on the minimum inscribed circle radius;

taking the coordinates of pixel points in the region outside the region to be rotated in the image to be transformed as static coordinates;

determining pixel coordinates after angle transformation according to the transformation matrix and the pixel point coordinates in the region to be rotated;

and obtaining an image set according to the static coordinate and the pixel coordinate after the angle transformation.

5. The method of claim 1, wherein determining a dual threshold from the thin-edged image set after the thin-edged processing of the image edges determined based on the image set comprises:

carrying out direction difference on each image in the image set to obtain a direction difference image;

carrying out non-maximum suppression processing on the direction difference image to obtain an image set subjected to thin edge processing;

determining pixel grading according to the image set subjected to the thin edge processing;

determining a dual threshold from the pixel hierarchy.

6. The method of claim 5, wherein determining a pixel rank from the set of edge-processed images, determining a dual threshold from the pixel rank, comprises:

grading the pixel values of the images in the image set after the thin edge processing according to the image bits;

dividing the classified pixel values into a non-edge point pixel value class, a weak edge point pixel value class and a strong edge point pixel value class according to a first threshold value to be solved and a second threshold value to be solved;

constructing a non-edge point pixel value class variance, a weak edge point pixel value class variance and a strong edge point pixel value class variance according to the classified pixel values, the to-be-solved first threshold and the to-be-solved second threshold to obtain an intra-class variance group;

establishing a minimum evaluation function according to the intra-class variance group, and solving the deviation of the minimum evaluation function to obtain a minimum calculation formula;

traversing the first threshold value to be solved and the second threshold value to be solved in the classified pixel values to obtain the solved first threshold value and the solved second threshold value which meet the minimization calculation formula.

7. The method according to claim 1, wherein the dividing pixel points of each image in the image set into a strong edge of the image, a weak edge of the image, and a non-edge of the image according to the dual threshold to obtain an image set after edge extraction comprises:

determining pixel points higher than a first threshold in each image in the image set as strong edges, determining pixel points higher than a second threshold and lower than the first threshold as weak edges, and determining pixel points lower than the second threshold as non-edges;

and performing edge extraction on the image set based on the strong edge, the weak edge and the non-edge to obtain an image set after edge extraction.

8. The method of claim 7, further comprising, after obtaining the weak edge:

determining whether each pixel point in the region formed by the weak edge has a strong edge pixel point in a preset numerical neighborhood;

if yes, setting the strong edge pixel point as a strong edge;

and if not, removing the pixel points.

9. An image edge feature library construction device, comprising:

the image transformation module is used for carrying out scale and/or angle transformation on at least one edge image to be extracted to obtain an image set, wherein the image set comprises multi-angle images and/or multi-scale images;

the determining module is used for performing edge thinning processing on the image edge determined based on the image set and then determining a double threshold value according to the image set after the edge thinning processing; wherein the dual threshold comprises a first threshold and a second threshold, the first threshold being greater than the second threshold;

the dividing module is used for dividing pixel points of each image in the image set into a strong edge of the image, a weak edge of the image and a non-edge of the image according to the double thresholds to obtain an image set after edge extraction;

and the construction module is used for constructing an image edge feature library according to the scale levels of the images in the image set after the edge extraction so as to extract the image edges.

10. A computer device, comprising:

one or more processors;

storage means for storing one or more programs;

the one or more programs are executable by the one or more processors to cause the one or more processors to perform the image edge feature library construction method of any of claims 1-8.

11. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, implements the image edge feature library construction method according to any one of claims 1 to 8.

Images